Inspiration
[cite_start]Tyto-Shield was inspired by the critical bottlenecks faced during earthquake and structural collapse rescue operations, where heavy dust, dense smoke, and steel-reinforced concrete render standard wireless rovers and cameras completely useless[cite: 4, 6, 8]. [cite_start]We looked to nature to solve these challenges, engineering a multi-biomimetic system: an invertible tracked chassis for unturnable locomotion [cite: 14, 15][cite_start], a polycarbonate curved turtle-shell carapace for impact deflection [cite: 16][cite_start], and an asymmetrical digital microphone array mimicking a barn owl's hyper-sensitive hearing to locate trapped survivors under ruins entirely through sound[cite: 18].
What it does
[cite_start]The subsurface system acts as an autonomous locator that navigates harsh debris[cite: 11, 12]. [cite_start]It processes the Time Difference of Arrival (TDOA) of sound waves to dynamically steer toward vocal distress or rhythmic tapping until it reaches the source[cite: 25, 26]. [cite_start]Once there, it stabilizes to gather audio logs via Azure AI Speech, reads ambient hazard risks (Carbon Monoxide and Methane) through electrochemical sensors, and generates optimal extraction strategies for rescue teams using an Azure OpenAI (GPT-4o) co-pilot dashboard[cite: 39, 40, 43, 45, 48].
How we built it
[cite_start]We engineered a scaled-down proof-of-concept prototype built around the dual-core ESP32-S3 MCU, programmed in native C/C++ running FreeRTOS for strict real-time multi-task scheduling[cite: 20, 21].
- [cite_start]The Firmware & Hardware: Implements an Extended Kalman Filter (EKF) fusing 6-DOF IMU data and high-resolution rotary encoders to minimize continuous track slippage[cite: 28, 29]. [cite_start]An onboard micro-motorized system actively spools out a protective Kevlar-braided communication tether matching the robot's velocity to prevent snagging[cite: 31, 32, 33].
- The Cloud Integration: Real-time data packages are securely streamed through an enterprise Azure IoT Hub pipeline.
- The Dashboard: Deployed via Azure Static Web Apps (Free Tier), displaying a live 3D navigation trajectory route utilizing Azure Maps alongside unified telemetry gauges powered by Power BI Embedded.
Challenges we faced
- [cite_start]Dead Reckoning Drift: Track slippage under unstable rubble accumulates coordinate calculation errors[cite: 27]. [cite_start]We mitigated this by writing custom EKF firmware and implementing an Acoustic Peak Intensity Recalibration routine that uses the physical sound source as a spatial anchor point to clear drift once the target is reached[cite: 28, 30].
- Cloud Dependency in Blackouts: Infrastructure failures mean sudden internet cutoffs. We solved this by implementing an Offline Fallback Mode, training a localized sound-classification and sensor-processing TinyML model via Edge Impulse to run directly on the edge MCU if cloud connection drops.
- Environmental Noise Isolation: Heavy site drilling can blind acoustic sensors. [cite_start]We integrated adaptive bandpass frequency filtering locally and leveraged Azure AI Speech advanced audio noise suppression to isolate human frequencies from industrial background noise[cite: 40].
What we learned
We gained deep expertise in building hybrid edge-cloud systems, tuning embedded multi-sensor fusion filters (EKF), deploying resource-optimized TinyML models on resource-constrained microcontrollers, and designing enterprise-grade IoT streaming data architectures with Microsoft Azure.
Log in or sign up for Devpost to join the conversation.